LiveScribe turns live audio into readable, searchable text the instant it's spoken — a lectern mic, a panel, a talkback channel, each its own source at once. The speech model runs on your Mac. Nothing is streamed to a cloud speech vendor, ever.
Caption a sermon live for hard-of-hearing seating, an overflow room, or a printed order of service that goes off-script.
Caption dialogue or narration without adding a captioner to the run of show.
One feed per panelist mic, so a multi-speaker session stays legible instead of blurring into one caption stream.
Procedural clarity for the room, entirely local — no audio leaves the building over a network line.
Feed in a talkback or intercom channel and get a running text log of every call — so nobody has to ask what was just said on com.
Pair with the cloud viewer so a remote producer or an off-site interpreter follows the room live.
LiveScribe is built on Apple's on-device SpeechAnalyzer and SpeechTranscriber — the long-form, continuous transcription engine for macOS 26, not the old one-minute SFSpeechRecognizer API. There's no cloud speech vendor in the loop, no network dependency for the core feature, and no audio uploaded anywhere. The model downloads once on first run, then runs entirely offline.
Speech never leaves the machine to be transcribed. No per-minute API bill, and no vendor outage that takes your captions down mid-show.
Once the on-device model is downloaded, transcription keeps running even if the venue's Wi-Fi doesn't.
SpeechAnalyzer is the long-form API — designed to run for the length of a service or a show, not chunked into one-minute calls.
Every source is its own channel — its own input device, its own channel number, its own label, icon and color. Pick Dante Virtual Soundcard, channel 5 of 64, and only that signal gets transcribed.
A talkback or intercom channel is just another source — pick the device and channel like any mic, and cue calls, standbys, and crew chatter become a live, readable log instead of something everyone has to half-remember or ask to repeat.
A recreation of the real unified feed, not a screenshot — each channel keeps its own color, the line still being spoken stays dimmed until it lands, and text size is adjustable for reading from across the room.
LiveScribe can open a view-only room on a small cloud relay and hand out a join code. The Mac makes an outbound connection only — it never opens an inbound port — and nothing flows back from a viewer. Useful for a remote producer, an overflow room, or a booth that just needs the captions without the gear.
These aren't shipped yet — they're next up, in roughly this order.
Apple's on-device Translation framework, applied per channel — still local, still no cloud vendor.
Keyword rules that fire OSC over UDP or MIDI, off the keyword-monitoring seam already built into the engine.
Sync transcripts across machines over Bonjour + UDP — no relay needed when everyone's on the same network.
Save and load a full channel + automation setup as JSON, so a rig's config travels with the show file.
LiveScribe is one app among several — each native to macOS, each built by people who actually run live shows.
Do this when that happens — a rule engine that watches your console, wireless and playback.
Broadcast clocks and countdowns, fanned to display, web, NDI and SDI at once.
A video matrix router for NDI, DeckLink, SRT and web feeds.
Multi-source audio listening and cueing across sources, buses and outputs.
Real-time, on-device captions for live production. You are here.
LiveScribe is in active development. No cloud speech bill, no round trip to a server — just a Mac listening to the room.